Research On Maneuvering Target Tracking Algorithm Under One-step Measurement Random Delay

Maneuvering target tracking is a critical research direction in the field of target tracking,which has been widely used in both military and civilian applications.The traditional maneuvering target tracking method assumes that measurement information can arrive in realtime at the processing center of the tracking system,and that this information can be used to correct predicted results.However,in practical applications,measurement information can easily be delayed due to communication bandwidth or network delay reasons,and as a result,the tracking system may not obtain real-time measurement information.Therefore,traditional methods are no longer applicable.In addition,the traditional interacting multiple model algorithms with one-step measurement random delay suffer from problems such as low model switching efficiency and lag in switching.This thesis proposes a Bayesian filtering framework for one-step measurement delay and an adaptive interacting multiple model algorithm for one-step measurement random delay to address the aforementioned issues.The specific details are as follows:(1)A Bayesian filtering framework with one-step measurement random delay is proposed for the maneuvering target tracking problem under one-step measurement random delay.Firstly,the impact of noise on target state estimation under measurement delay is analyzed.The measurement noise vector is introduced into the target state vector,and a new augmented state vector is obtained.Secondly,based on the Gaussian assumption under standard state,two Gaussian assumptions under measurement random delay are proposed.Based on the standard Bayesian filtering framework and Gaussian assumptions,a Bayesian filtering framework for onestep measurement random delay is proposed.This filtering framework divides the entire filtering process into two parts:the measurement noise filtering estimation and the state filtering estimation.The Bayesian filtering framework for one-step measurement random delay is applied to the EKF,UKF,and CKF algorithms,and algorithms corresponding to one-step measurement random delay are obtained.Finally,simulation experiments are conducted for verification.The results indicate that the proposed method improves the real-time performance and accuracy compared with the standard nonlinear Kalman filtering under the condition of one-step random measurement delay.(2)An adaptive interacting multiple model algorithm for one-step measurement random delay is proposed to address the problem of inefficient IMM algorithm model switching under one-step measurement random delay.Firstly,this paper investigates the IMM algorithm,analyzes its advantages and disadvantages,and discusses the impact of TPM on the IMM algorithm.It concludes that the TPM should be updated in real-time with the target motion state.Secondly,in order to realize the dynamic update of TPM,a correction function is proposed by introducing the current model information and the past model information,and the TPM is corrected by the correction function,so as to improve the efficiency of switching between models and make the algorithmic model match the real model better.Finally,the validation is carried out by simulation experiments.The results showed that the proposed method has higher tracking accuracy and better real-time performance than the standard IMM algorithm under the condition of one-step measurement of random delay.(3)The two proposed algorithms are implemented using Python and PyQt5 in this study.The software design,development and testing environment are described from a software engineering perspective.Ultimately,the software achieved its initial design goals and reduced the operational threshold for users who are unfamiliar with tracking maneuvering targets under one-step measurement random delay.In summary,this thesis analyzes the existing problems in maneuvering target tracking under one-step measurement random delay,and proposes a filtering framework and an adaptive interacting multiple model algorithm under one-step measurement random delay.The proposed algorithms have better accuracy and real-time performance compared to traditional algorithms under one-step measurement random delay.